The Archive

Qwen 3.6 27B quantization benchmark: BF16 (69.78% avg accuracy, 54GB RAM) vs Q4_K_M (66.54%, 15GB) vs Q8_0 across HumanEval, HellaSwag, BFCL.

Qwen 3.6-27B achieves 38.2% on Terminal-Bench 2.0 coding tasks, demonstrating open-weight models viable for real-world agent work.

GPT-5.5 ranks 2nd on Extended NYT Connections Benchmark; Kimi K2.6 leads open-weights; DeepSeek V4 Pro and Qwen 3.6 Max show significant gains.

I'd jump on runpod and ssh in to test my workloads, but they don't have it. Would love to know how well this runs, particularly as context approaches a full 256K. Thanks!

User reports GBNF grammar optimizations for Qwen 3.6 35B and 27B models improving coding task performance in llama.cpp.

User reports Qwen 3.6 27B outperforms 35B variant for local coding tasks on consumer hardware.

User reports Qwen 3.6 35B quantization variant (A3B Heretic) performs well locally with low KLD divergence and multi-turn capability.

Developer documents practical setup for running Qwen 3.6 35B on M2 MacBook Pro 32GB via llama.cpp, with performance notes and optimization tips.

Reddit user claims Qwen 3.6 27B outperforms Claude Sonnet 4.6 on planning tasks in direct comparison.

KL divergence benchmark comparing Gemma 4 and Qwen 3.6 with quantized KV cache compression (q8_0, q4_0).

Qwen 3.6 35B-A3B MoE model achieves 250+ tok/s on AMD Radeon 780M iGPU via llama.cpp Vulkan.

User reports Qwen 3.6 27B quantized model achieves 22 tok/s on RTX 5060 Ti with 24K context window.

Reddit user benchmarks Qwen 3.6 35B vs 27B on coding tasks; 35B faster (72 TPS) but less accurate, 27B slower but more precise.

Qwen 3.6 27B matches Claude Sonnet 4.6 on agentic benchmarks, surpassing GPT 5.2/5.3 and Gemini 3.1 Pro.

Technical walkthrough: Qwen 3.6 27B achieves 85 TPS, 125K context on single RTX 3090 using llama.cpp.

User reports success running Qwen 3.6 35B with local coding agent using structured planning skill file on production code.

User reports Qwen 3.6 27B practical performance on data science and Python debugging tasks with 24GB VRAM.

Performance test: Qwen 3.6 27B with speculative decoding achieves 25.53 tokens/sec with 2x speedup on local hardware.

User reports Qwen3.6-27B exhibits strong agentic behavior and willingness to iterate on coding tasks in local deployment.

User demonstrates Qwen3.6-27B running locally via llama-server with 200k context on dual RTX 3090, achieving coding performance cheaper than Claude.

Qwen3.6-27B model release announcement on Hugging Face with no additional details.

Interactive benchmark comparing coding capabilities: Qwen 3.6 35B, Qwen 3.5 variants, Gemma 4, GLM 4.7 Flash via racing game simulation.

Developer switching from Claude Opus 4.7 to Kimi K2.6 citing performance degradation and cost, supplementing with Qwen 3.6.

User comparison of Gemma 4 26B and Qwen 3.6 35B performance on consumer GPU with anecdotal quality assessments.